When I load a batch of MCM to process, I see intermittently, the run time jump from the normal 2-3 hours per task to over 1 & sometimes over 2 days. They actually do run for that length of time. Machine is nowhere near maxed out. Any thoughts?

One thing to check: earlier this year, I had a similar problem with a newly purchased REFURBISHED MFF[1] unit and ALL WU’s I was attempting to process.

It seems that the provided AC adapter was not DESIGNED to be used with that system and because of that never was able to communicate with CMOS. So, the system ALWAYS was fully throttled back (until the proper AC Adapter was provided).

A quick check would be to pull up the TASK MANAGER, select the performance tab, then select CPU - and see if the indicated CPU speed is as expected -or- significantly lower. Then proceed from there.

Happy crunching.

1. MFF - Micro Form Factor

The other thing you may want to look at is the power settings in the operating system.
The other thought is if your system has some performance cores and some efficiency cores. The performance cores will show greater throughput than the efficiency cores.
Cheers

Thank you. Very insightful.

Thank you, Sgt. Joe - good diagnostic suggestions!... Instead of posting my [overlong] attempt on the same topic, I'd just like to identify some "fine details" it might be interesting to know:

• How is the user's BOINC client set up -- % of cores and % of CPU in particular[1]
  
• what is the user's work mix when problems occur -- all MCM1?
  
• Is the MCM1 task VMethod=LOO or VMethod=NFCV?[2] - (in the result report on the web site);
  
• Is the user using other software on the system when problems occur.

Unfortunately, without access to the MCM1 source code and build environment all we can do is try to gather information and see where that takes us -- given that MCM is in its home territory here, that might well help them do something about the problems as and when they eventually have time. And I'd happily take part in a proper Beta test for a revised version of MCM1 :-)

Cheers - Al.

[1] The "% of CPU time" value seems to have played a part in issues with MCM1 tasks looping during initialization. If that's a timer problem, it might also affect this other issue.

[2] I mention the VMethod thing because [as you may recall] discussions about why there seem to be two distinct sorts of run-time for MCM1 seemed to suggest that VMethod=NFCV tasks spent a lot longer calling various O/S features.